Evaluation of broken steel wire rope using magnetic flux leakage and optimised convolutional neural network

Wire rope has become a common component in mines because of its advantages such as high strength and flexibility. Nevertheless, prolonged utilisation in severe environments can result in damage due to breakage. To address this problem, this paper proposes a damage quantification method for multi-objective optimised convolutional neural networks incorporating tiny damage signal enhancement. Firstly, the signal characteristics of wire rope breakage damage are analysed by constructing a magnetic dipole model and exploring the applicability of axial and radial signals for quantitative identification. Secondly, an algorithm to enhance the radial damage signal of a small broken wire is proposed, which can enhance the peak-to-peak value of 1 broken wire by about 2.4 times, and improve the detection ability of small broken wire damage. Finally, a convolutional neural network (CNN) was constructed, and the structural parameters of the network were optimised using the crown porcupine optimisation (CPO) algorithm for quantifying wire rope breakage damage. The experimental results show that the model proposed in this study exhibits excellent performance in terms of accuracy, precision, recall, and F1 score compared to multiple models. The accuracy is augmented to 99.17%, thereby enhancing the quantisation precision of broken wire defects.